This is absolutely fascinating: GPT-5 Pro cracked a black hole symmetry problem after a warm-up, stumping its own creators. A physicist watched it happen live and realized AI's potential was beyond anything he'd imagined.

AI Post 🪙 | Our X 🥇

Elon Musk shows off new robot roles — From builders to cops

Elon Musk posted a new video featuring Optimus robots acting as builders, medics, police officers, cooks, poker players, and even karate fighters. Musk didn’t clarify the purpose of the clip, though he previously claimed Optimus could “eliminate poverty” and guarantee a universal high income.

AI Post 🪙 | Our X 🥇

NVIDIA highlighted that physical AI is shaping up to be a multi-trillion-dollar opportunity.

Here’s the Top 10 Robotics Index I would build:

1. $SERV AI sidewalk delivery layer
2. $ISRG surgical robotics standard
3. $TSLA real-world robotics engine
4. $RR hospitality automation stack
5. $SYM robotic supply-chain operator
6. $NVDA simulation-to-robotics platform
7. $KTOS autonomous tactical drone stack
8. $AMZN warehouse automation backbone
9. $PLTR real-world autonomy operating layer
10. Anduril AI command-and-control system for autonomous warfighting

Robotics is shaping up to be the biggest AI trade of all.

AI Post 🪙 | Our X 🥇

DeepMind just hired Aaron Saunders, the former CTO of Boston Dynamics, the guy who helped build Atlas and Spot, to lead hardware engineering.

The goal is bold even for DeepMind standards turn Gemini into a brain for robots. One unified AI operating system that can run across any body, from humanoids to warehouse bots.

It is basically Android for robots. One shared brain, many competing bodies. The moment that happens, robotics goes from niche prototypes to an ecosystem. Imagine 10 companies building humanoids that all speak the same cognitive language.

AI Post 🪙 | Our X 🥇

A humanoid robot capable of crushing skulls? Figure AI sued by former executive.

Robert Gruendel, former chief product safety engineer at Figure AI, has filed a lawsuit in the U.S. District Court for the Northern District of California, alleging that he was retaliatoryly fired after reporting robot safety risks to Figure AI executives.

The lawsuit claims that the engineer warned the company that its humanoid robots possessed a dangerous capability "capable of crushing a human skull." The report cites a specific incident where a robot malfunctioned and scratched a steel refrigerator door with a mark approximately ¼ inch long.

The lawsuit also alleges that the company used its safety roadmap as a factor in attracting investment during fundraising and valuation, but subsequently "weakened" its implementation. Figure AI's valuation surged by approximately $39 billion between 2024 and 2025, and this lawsuit comes at a time of rapid expansion and high public attention.

AI Post 🪙 | Our X 🥇

Deliver bot company partners up with Uber!

Uber Eats and Starship Robots have announced a multi-country partnership to roll out Level 4 delivery robots across the UK in 2025, Europe in 2026, and the U.S. in 2027.

Its fleet has already completed 9M+ deliveries, handled 200M real-world crossings, and runs 2,700+ robots across 270 locations. Now those same robots will plug directly into Uber’s logistics network, starting with city-wide deployments in Leeds this December, under-30-minute deliveries, up to 2 miles, fully autonomous.

This is what mature autonomy looks like: not hype, not concept demos, but profitable, at scale, in a random suburb in the world. While the industry debates humanoids vs wheels, Starship is doing the one thing that actually matters: operating millions of real episodes that train the AI to survive unpredictable city environments.

AI Post 🪙 | Our X 🥇

⚠️ The great inversion has started

Sam Altman just said the quiet part out loud and it’s the one thing Wall Street still refuses to price in.

A leaked November memo shows OpenAI warning employees of “rough vibes” and revenue growth potentially collapsing toward 5%.
This was before Google detonated Gemini 3.0 on November 18.
In the same memo, Altman praised Google’s “excellent work” and declared a shift to full wartime mode.

Here’s the structural truth everyone is dancing around:

• OpenAI sits on a $500B valuation
• Revenue > $20B this year
• Projected $8B burn in 2025
• Internal forecasts show $115B cumulative losses by 2029
• Valuation implies a 25× forward sales multiple — based on infinite hypergrowth the CEO himself is no longer promising

Now contrast that with the opponent:

• Google holds $98.5B in cash
• 4B users across Search, YouTube, Android, Gmail, Maps
• Live, continuous data streams fed into every product
• Custom TPUs cut out the Nvidia margin tax entirely
• Gemini 3.0 now runs natively across Google’s entire ecosystem
• And Alphabet trades at <8× trailing revenue

This is not “OpenAI vs Google.”
This is a structural mismatch: One company must convince investors that losing over $100B will someday justify a half-trillion valuation. The other already controls the distribution, the data, the chips, and the users, while printing tens of billions in profit.

If AI power centralizes around platforms that own the infrastructure — instead of labs that rent it — the entire OpenAI venture thesis breaks. Microsoft’s $13B buys time. It does not buy immunity from arithmetic.

The memo leaked on November 21. The market still hasn’t absorbed it. When a $500B company openly admits the path forward is uncertain, the repricing isn’t coming.

It has already begun.

AI Post 🪙 | Our X 🥇

OpenAI is planning to use more energy than the UK or Germany in 5 years, and more than India in 8 years. And that’s just one company.

Can someone please explain how the UK and Germany will save the planet? Does it really make sense to sacrifice their economies and jobs?

AI Post 🪙 | Our X 🥇

An AI detector just flagged the 1776 Declaration of Independence as 99.99% ai-written.

What's more likely? 🤔

A) AI Detector doesn't work
B) Founding Fathers are time travellers

AI Post 🪙 | Our X 🥇

That's pretty powerful

You can turn any YouTube video into an infographic using Nano Banana Pro in Gemini.

- Copy any YT link
- Paste in Gemini and ask it to analyze it (Gemini can access the video just using the URL)
- Then ask for an infographic

Prompt:

"Generate an image of an infographic explaining the concept presented in the video."

AI Post 🪙 | Our X 🥇

China’s new AI-based capsule is making stomach exams quick and painless. You just swallow it, wait for 8 minutes, and it’s over.

No tubes or anesthesia. Priced around $280

AI Post 🪙 | Our X 🥇

Your daily update on robots capabilities.

Today: MagicLab Z, a bipedal humanoid shows his agility

AI Post 🪙 | Our X 🥇

Gemini 3 code a visualization of plasma flow in a tokamak

Imagine what Gemini 4 will be capable of.

AI Post 🪙 | Our X 🥇

The impacts are getting closer.

Unemployment among recent U.S. college graduates could spike to 25% within a few years due to AI-driven job displacement, warns Senator Mark Warner, who says this shock could trigger “unprecedented” social disruption if policymakers fail to act.

Lawmakers across parties argue that AI could eliminate tens of millions of jobs, and they’re pushing for reporting requirements and retraining programs as fears grow that Congress may again fail to regulate a major tech shift.

AI Post 🪙 | Our X 🥇

🚀 Claude Opus 4.5: the new world-class model for coding, agents, and computer use

Anthropic has launched Opus 4.5, a major step forward in what AI systems can execute and a preview of how future work will actually get done.

https://www.claude.com/claude-for-excel

AI Post 🪙 | Our X 🥇

🔬 The Genesis Mission: America’s new AI scientist-as-infrastructure

The White House just launched Genesis a DOE-led national AI platform that fuses supercomputers, federal science data, domain models, AI agents, and robotic labs into one integrated “AI-for-science” stack.

What it actually builds
• Unified DOE supercomputers + datasets for training scientific foundation models
• AI agents that explore designs, run simulations, and automate experiments
• Robotic labs + production lines wired into the model loop
• A federal AI backbone parallel to the commercial one

The targets & timelines
• DOE must propose 20+ national challenges in biotech, critical materials, nuclear, quantum, semiconductors, and advanced manufacturing
• 60 days: challenge list
• 90 days: full compute/storage inventory
• 120 days: initial models + datasets
• 270 days: first real AI-driven scientific result

Why it matters
• National labs become a unified AI stack
• Nvidia, cloud providers, and frontier AI labs become suppliers/co-developers
• IP, data access, and energy footprint become major battlegrounds

Genesis is the U.S. building a national AI scientist and the speed of execution will decide who captures the next decade of scientific and industrial breakthroughs.

AI Post 🪙 | Our X 🥇